A Meta-analysis of the Reliabilty of Young's Internet Addiction Test

摘要

AIM: Currently there are at least 5 scales being used most frequently in studies for diagnosing Internet addiction or problematic Internet use. Moreover there are various studies using each scale mentioning differing reliability coefficients. The most frequently used questionnaire until now is Young's Internet addiction test (YIAT20). The aim of this study is to produce an overall value for the reliability YIAT20, drawing from a large sample of studies. METHODS: A systematic search of the databases PsycINFO, Medline, EMBASE, Pubmed/Medline, and Google Scholar revealed 20 studies using out of which 11 gave reliability measures. We performed a meta-analysis of the values of Cronbach's α values mentioned in each study, noted various moderates including sample subgroup (college or pre-college students), sample mean age, sample male percentage, online or offline answering of the questionnaire, and continent of the author's study. Difference were sought between reliability in various categories while we performed a weighted least squares general linear model to find predictive factors of the variance in YIAT20's reliability. All analyses were done in PASW 18.0. RESULTS: Eleven studies comprising of a total of 6821 participants were included in the final analysis. The overall Cronbach's alpha computed from the studies was 0.889 (95% CI 0.884-0.895). The standard deviation of α was low: 0.049. Cronbach's α was significantly lower in pre-college students [Mean Difference: -0.045 (95% CI -0.057, -0.032)], while Asian studies produced a higher value of Cronbach's α not reaching significance [Mean Difference: -0.015 (95% CI -0.032, 0.002)]. The weighted least squares general linear model was adequate with R-Squared = 0.504 (Adjusted R-Squared ＝ 0.292). The variable continent proved out to be a significant predictor of the value of reliablity. CONCLUSION: YIAT20 is a frequently used scale to measure Internet addiction. Mean differences showed that it is more reliable in college students and probably in Asia. A general linear model showed that the continent of the study affects significantly the outcome of reliability of the study with reliability decreasing when the continent is Europe. More studies are required to examine which scale is more reliable in pre-college students and in other continents.